Ask Claude What It's Thinking, It Reports J-space
dmvaldman · x · 2026-07-07
A repost points out that when you ask Claude what it is "thinking," it reports the contents of its "J-space," and you can alter its choices within that space. This relates to recent research on model internal representations and interpretability.
This post shares the same theme as concurrent Jacobian lens discussions, so no separate notification tag is applied here.
Related event: Anthropic Discovers Global Workspace Inside Claude(102 posts)→
More from Research
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Terence Tao's SAIR Foundation revamps Mathathon with $1.5M compute per team — _sathvikr · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27